Arch Linux

Recursive Pineapple

Member

Registered: 2024-10-30

Posts: 3

[SOLVED] Nvidia drivers aren't working

My 3060ti's drivers have been broken for the past few weeks. I've tried several things to get them working again, and nothing has fixed the problem. They work 90% of the time, but I know there's a driver error because I consistently get a SIGBUS crash when I close certain games. I also can't record anything with any screen capturing software because the hardware encoder is malfunctioning. nvidia-settings also shows almost no info. I can see the basic info for my GPU, but nothing else. nvidia-smi/nvtop shows the expected output, so the drivers are installed and mostly-working. When I start games with __GL_THREADED_OPTIMIZATIONS=0, they close properly.

I upgraded the drivers to the version in the testing repo, which fixed the hardware encoder somewhat (though it still refuses to encode quickly), but everything else is still broken.
The nvidia driver download page says the latest drivers for the 3060ti is 550.127.05. I tried downgrading to 550.90.07 by downloading the packages from the archive and installing them manually, but I can't install them because of the following error. I'm not really sure how to fix this without breaking my system.

❯ sudo pacman -d -U nvidia-550.90.07-1-x86_64.pkg.tar.zst nvidia-utils-550.90.07-1-x86_64.pkg.tar.zst 
[sudo] password for [redacted]: 
loading packages...
warning: downgrading package nvidia (565.57.01-1 => 550.90.07-1)
warning: downgrading package nvidia-utils (565.57.01-1 => 550.90.07-1)
resolving dependencies...
looking for conflicting packages...

Packages (2) nvidia-550.90.07-1  nvidia-utils-550.90.07-1

Total Installed Size:   634.66 MiB
Net Upgrade Size:      -108.86 MiB

:: Proceed with installation? [Y/n] y
(2/2) checking keys in keyring                                                                                                  [#############################################################################] 100%
(2/2) checking package integrity                                                                                                [#############################################################################] 100%
(2/2) loading package files                                                                                                     [#############################################################################] 100%
(2/2) checking for file conflicts                                                                                               [#############################################################################] 100%
error: failed to commit transaction (conflicting files)
nvidia-utils: /usr/lib/libnvidia-egl-gbm.so exists in filesystem (owned by egl-gbm)
nvidia-utils: /usr/lib/libnvidia-egl-gbm.so.1 exists in filesystem (owned by egl-gbm)
nvidia-utils: /usr/share/egl/egl_external_platform.d/15_nvidia_gbm.json exists in filesystem (owned by egl-gbm)
Errors occurred, no packages were upgraded.

This is the error I get when trying to close a game. The java stacktrace shows that this is thrown while destroying the window (glfwDestroyWindow).

# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGBUS (0x7) at pc=0x000075499625163c, pid=8519, tid=8522
#
# JRE version: OpenJDK Runtime Environment JBR-21.0.4+8-569.1-jcef (21.0.4+8) (build 21.0.4+8-b569.1)
# Java VM: OpenJDK 64-Bit Server VM JBR-21.0.4+8-569.1-jcef (21.0.4+8-b569.1, mixed mode, sharing, tiered, compressed oops, compressed class ptrs, g1 gc, linux-amd64)
# Problematic frame:
# C  [libnvidia-glcore.so.565.57.01+0xa5163c]
#
# Core dump will be written. Default location: Core dumps may be processed with "/usr/lib/systemd/systemd-coredump %P %u %g %s %t %c %h" (or dumping to /home/[redacted]/Documents/minecraft/GT5-Unofficial/run/core.8519)

Is there a way to figure out what the problem is? This started happening immediately after 'pacman -Su' updated the drivers so I'm guessing the problem is the driver version, but I'm not sure how to verify that. My GPU isn't that old so I'd be surprised if this was the case.
I don't see any useful errors in dmesg, Xorg.0.log, or 'journalctl -b'. I'm not sure what other logs I should check. The only error I see is from the java coredump, but it doesn't show any function names in the stack.

Last edited by Recursive Pineapple (2024-10-30 21:14:31)

ptr1337

Member

Registered: 2024-09-24

Posts: 10

Re: [SOLVED] Nvidia drivers aren't working

NVIDIA has different driver branches:

New Features (560/565), Stable (550 and 535).

You wont be able to downgrade to the 550 driver, which was provided in the past by archlinux, because this one is outdated as well as does not contain the recent changes to compile properly against the linux kernel.
Either you manually adjust the PKGBUILD or use the nvidia-550xx-dkms package from the AUR, but I think this one is missing lib32 packages.

Otherwise, you can also downgrade to the 560 driver, but since you had issues already before, this wont likely fix it.

You can also downgrade to the 56
You can use these for now:
https://github.com/CachyOS/CachyOS-PKGB … idia-550xx

Check if the issue is also present there, otherwise report the issue to nvidia.

Recursive Pineapple

Member

Registered: 2024-10-30

Posts: 3

Re: [SOLVED] Nvidia drivers aren't working

The packages you linked won't install because of the same egl-gbm conflict.

error: failed to commit transaction (conflicting files)
nvidia-550xx-utils: /usr/lib/libnvidia-egl-gbm.so exists in filesystem (owned by egl-gbm)
nvidia-550xx-utils: /usr/lib/libnvidia-egl-gbm.so.1 exists in filesystem (owned by egl-gbm)
nvidia-550xx-utils: /usr/share/egl/egl_external_platform.d/15_nvidia_gbm.json exists in filesystem (owned by egl-gbm)
Errors occurred, no packages were upgraded.

It sounds like this is a driver bug so I'll see if nvidia can help, but if you have any other ideas I'll try them. I'm starting to think the firmware on my GPU got corrupted somehow.

seth

Member

Registered: 2012-09-03

Posts: 58,807

Re: [SOLVED] Nvidia drivers aren't working

That's not a driver bug, https://archlinux.org/packages/extra/x86_64/egl-gbm/ was introduced a month ago and is a dependency of the 560xx versions of nvidia-utils, but the 550xx packages had that file.

I don't buy that the SIGBUS is because of some driver issue.
=> restore the status quo ante and

I don't see any useful errors in dmesg, Xorg.0.log, or 'journalctl -b'. I'm not sure what other logs I should check. The only error I see is from the java coredump, but it doesn't show any function names in the stack.

post all of them, also because

. I also can't record anything with any screen capturing software because the hardware encoder is malfunctioning. nvidia-settings also shows almost no info. I can see the basic info for my GPU, but nothing else.

suggests you're not actually running on the nvidia GPU or you're running wayland.

---

How to upload text · How to boot w/o GUI · Disable Windows Fast-Start! · Fix your xinitrc

Recursive Pineapple

Member

Registered: 2024-10-30

Posts: 3

Re: [SOLVED] Nvidia drivers aren't working

You're right! Somehow GDM started using Wayland instead of X11. I disabled Wayland and everything is fixed now. It's super bizarre that it manifested in all of those ways because it even affected the H264 CPU software encoder in OBS. I should've clued in when I noticed that the Xorg log looked like it was mangled (it only had a few seconds of logs around 50s).